Great. What I worry about however is linking with external libraries:

http://stackoverflow.com/questions/6494101/any-issues-with-mixing-libraries-with-and-without-std-c0x

We might be ok but I have no idea.

Martin
On 2014-01-29 15:56, Anders Logg wrote:

> On Wed, Jan 29, 2014 at 03:31:10PM +0000, Garth N. Wells wrote:
>
>> A long thread has been mulling over the build system consequences of
>> merging FFC and UFC into one 'package':
>>
>>     http://fenicsproject.org/pipermail/fenics/2014-January/000896.html
>>
>> The crux of it is the different build systems used by FFC (Python
>> distutils) and UFC (CMake). UFC uses CMake because of the dependency
>> on boost::shared_ptr.
>>
>> If we were to switch to std::shared_ptr, using distutils for both
>> FFC and UFC should be straightforward. However, std::shared_ptr is
>> part of the C++11 standard and would require a compiler that
>> supports at least the std::shared_ptr part of the standard.
>>
>> The question therefore is should we switch from boost::shared_ptr to
>> std::shared_ptr in DOLFIN? To start things off:
>>
>> * Pros:
>>
>>  - Makes FFC and UFC packaging simple
>>  - Use system provided STL (shared_ptr, unordered_set,
>> unordered_set, etc)
>>  - We can start using other useful C++ features from the new
>> standard, e.g. auto:
>>
>>   for (std::vector<<<..............> > >::const_iterator itr =
>> myvec.cbegin(); itr != myvec.cend(); ++itr)
>>
>> becomes
>>
>>   for (auto itr = myvec.cbegin(); itr != myvec.cend(); ++itr)
>>
>>
>> * Cons
>>
>>   - Lack of C++11 support in old compilers. This is unlikely to be a
>> problem on desktops. Any issues are most likely to be on older
>> clusters. It does appear that std::shared_ptr has been available for
>> major compilers (GCC/Clang/Intel/IBM/Microsoft) for quite some
>> years.
>>
>
> I support switching to std::shared_ptr and look forward to using
> auto.
>
>
Ok. I'll drop some small C++11 'testers' in the code and we can see if they
cause any issues for users of dolfin/master.

Garth

 --
> Anders
>
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to